عليرضا اميدي گوكه

Dust in construction environments is known as one of the main health risks, which is caused by activities such as cutting materials, demolition operations, traffic of heavy machinery an‎d mining. These suspended particles can cause respiratory diseases, asthma an‎d other long-term complications in workers. To deal with this challenge, this research has developed an intelligent system based on image processing that is capable of detecting dust, recognizing people an‎d monitoring the condition of the mask in a cascade manner. To train the system models, a dedicated database was created, which includes 500 images of various dust production scenarios, 1000 images of people in real work environments, an‎d 820 balanced images of mask coverage status (masked, without mask, an‎d various types of safety covers). After training on the Google Kolb platform using the Python coding language, these models achieved an average accuracy of 76.1% in dust detection, 87.4% in person identification, an‎d 92.5% in mask coverage eva‎luation at the overlap threshold of 50%, an‎d then these models were used to create a cascade architecture. The practical eva‎luation of this system in 47 real situations of construction projects has confirmed its efficiency in diverse an‎d challenging environmental conditions. This system has the ability to be provided to the safety managers of construction projects to monitor high-risk environments in real time an‎d control the status of peopleʹs protection against dust by connecting to environmental cameras. This solution can play an effective role in reducing respiratory risks an‎d improving occupational health in construction workshops.

بينايي رايانه اي , شناسايي گرد و غبار , يادگيري عميق , ايمني كارگاهي , پايش ماسك

dust detection , computer vision , mask wearing monitoring , cascaded architecture

Alireza Omidi

Mostafa Khanzada